From: Paul H. Hargrove (PHHargrove_at_lbl_dot_gov)
Date: Tue Oct 04 2005 - 12:12:01 PDT
Hideyuki, Sorry about the website being down. We are in the process of recovering from an attack on our webserver. For now the project pages can be accessed at http://ftg.lbl.gov/CheckpointRestart/CheckpointRestart.shtml As to the usage question... The comments in include/libcr.h are meant to substitute for full documentation until we've had the chance to write it. However, feel free to ask questions if anything is not clear to you. You can find examples in the examples subdirectory of the blcr source directory. In general, an application with no speciallized resources to checkpoint (such as sockets) doesn't need any calls to blcr functions. It just needs to be linked w/ libcr.so, or it should be launched with cr_run (which LD_PRELOADs libcr.so). A checkpoint can then be requested using cr_checkpoint. If one needs to perform custom actions at checkpoint/restart time, then one should register a callback as can be seen in examples/pthread_misc/pthread_misc.c. It is in that callback that one would call cr_checkpoint(). If one wants an application to checkpoint itself at a given point in the execution, then one of the cr_request*() functions should be used. -Paul Hideyuki Jitsumoto wrote: >Sorry, I forgot to describe my environment. > >Machine: VMWare 5.00(x86) >OS: Linux 2.4.27 >GCC: 3.3.5 >BLCR: 0.4.0 > >And, I could use cr_run, cr_checkpoint and cr_restart command. > > >>Hello, >> >>I'm Hideyuki interested in MPI checkpointing. >> >>Now, I try to use BLCR from my program. >>But I don't get the specification of BLCR API, then >>I don't understand how to checkpoint my process. >> >>Would you tell me a paper about how to use BLCR API ? >>Already I tried following code, but a segmentation fault occured. >>I thought I have to use some API excluding cr_checkpoint()(Is this wrong??). >> >>---------------------------------------------------- >>main(){ >> while(){ >> sleep(1); >> ....... >> res = cr_checkpoint(0); >> if(res){ >> printf("restarted.\n"); >> }else{ >> printf("continued.\n"); >> } >> } >>} >>---------------------------------------------------- >> >>Finary, I found http://ftg.lbl.gov/ link was unabailable. >>All of the URI don't have ".", so I want to add "." to URI. >> >>I'm waiting for your advise, thank you. >>Sincerely yours, >>---------------------------------- >>Hideyuki Jitsumoto @ TITECH >>[email protected] >>---------------------------------- >> > > > > -- Paul H. Hargrove PHHargrove_at_lbl_dot_gov Future Technologies Group HPC Research Department Tel: +1-510-495-2352 Lawrence Berkeley National Laboratory Fax: +1-510-486-6900